Re: Rebooting first machine causes wackamole on second machine to abort

Hi,

I would recommend setting up Spread not to use multicast. It looks
like you have a multicast issue on your network.
Try this first and see if the problem persists:

Spread_Segment  128.243.40.255:4803 {
        haydn   128.243.40.92
        mozart  128.243.40.93
        mahler  128.243.40.94
}

Ciprian

DO> Hi

DO> I'm trying out wackamole to see if I could use it for HA of a
DO> mailing-list server address, testing on three identical machines
DO> (mahler, haydn and mozart), all running RedHat 8 (Linux 2.4.18). We
DO> currently use LVS, so I'd like to see if wackamole would be an
DO> alternative, particularly as I only want HA and not load balancing.
DO> Spread is installed and working OK on the three machines -- I can send
DO> messages between them with spuser. I'm using the local multicast address
DO> 239.0.0.1 for the machine group.

DO> I start wackamole on one of the machines successfully, assigning the
DO> virtual IP (128.243.40.99) to eth0:1. (However, it does log "No such
DO> interface" in /var/log/messages, as other people have reported in this
DO> list, even though the conf file only mentions eth0 -- it would be useful
DO> if the log message stated what interface it was objecting to).

DO> Starting wackamole on the second machine is OK as well -- for this
DO> exercise, I run it under strace. However, when I reboot the first
DO> machine, wackamole on the second machine dies with an ABRT signal. (To
DO> avoid sending the 31K of strace output to the list, I put it on our
DO> webserver at http://www.nottingham.ac.uk/~cczdao/wackamole/strace.txt)

DO> A similar thing happens using all three machines. I can start wackamole
DO> on the first two OK, but starting it on the third machine makes the
DO> instance on the second machine die with an ABRT signal.

DO> It's as if the group membership change when the first machine is
DO> rebooted, or when the third machine joins, upsets the second wackamole.
DO> For the three-machine case, it doesn't matter which order I start the
DO> wackamoles: it's always the second one which ABRTs, on whichever
DO> machine.

DO> spread.conf, wackamole.conf attached.

DO> David